Platform · Workflow Builder
A workflow builder that reaches past the API
Triggers, branches, loops and 700+ integrations like any visual builder — plus an agent step for judgment, a document step for files, approvals in Slack or Teams, and a bot step for the systems that have no API. One canvas, one run log.
Updated for the October 2026 release.
Agent Builder · invoice-agent.flow
Trigger — every weekday, 7:00
or webhook · or chat message
Read AP inbox
Outlook connector · unread with attachments
AI step — extract & decide
vendor, lines, totals → match against open POs
Human approval if > $5,000
Slack approval to controller · else continue
Post to QuickBooks
create bill · code GL · attach source
In one paragraph
The ElectroNeek Workflow Builder is the browser-based, no-code canvas of AI Workflows. A workflow starts from a trigger (schedule, webhook, app event, email, form, SFTP drop, chat or an MCP call), runs steps in order with branches (Router), loops and code steps, and can include four step types other builders don't have together: an AI agent with tools, an IntelliDocs document step, a human approval or task, and an RPA bot on a free Windows runner. Every run is logged step by step with timing; steps are tested from the canvas in one click; versions roll back. It is billed per run — not per step, not per user.
What it does — concretely
Triggers
Schedules, webhooks (with HMAC verification and binary files), polled or real-time app events, inbound email, public forms, SFTP, chat, manual, and an MCP tool trigger so another agent can start the flow.
Logic without code — and with it
Router with named branches, loops over lists and tables, formula and data functions for transformations, error handlers on any step, retries and continue-on-failure. A JavaScript/TypeScript code step with npm when a formula isn't enough.
Four steps nobody else combines
Run Agent (a goal, a model, tools), IntelliDocs (a document into fields), Approval or Assign Form / Wait for Task (a person decides), Run RPA Workflow (a bot in SAP GUI, AS/400, Citrix or a desktop app).
Testing and debugging
Test a single step from the canvas with a redesigned output view, a step timeline and a latency timeline per run, clear errors with a shortcut to the failed step, cancellable paused runs, deadlines on waiting steps, and failure emails with the step's details.
Versions, folders, variables
Draft and published versions with rollback; steps upgrade only when you choose; flows and tables in folders; shared encrypted Variables; JSON import and export. Connections show the account and its health.
Templates to start from
The Explore gallery holds ready-made workflows for common jobs; our curated use cases link to the matching template where one exists.
How it works
01
Pick a trigger
Where does the work start — an email, a schedule, a form, a file, a message, another agent? Connect the app once; connections are reusable.
02
Lay out the steps
Deterministic steps for the known path; an agent step for the judgment; a document step for files; a bot step for the screen. Put an approval before anything consequential.
03
Test, publish, watch
Test each step on real data from the canvas, publish a version, and read the run log the next morning. Tighten or loosen approvals as the history earns it.
What it doesn’t do (yet)
Stated here so you don’t find out in a pilot. Roadmap items are marked as such.
- —Run data is retained for 30 days; a flow paused on an approval cannot resume after that.
- —Tables are capped at 10,000 rows × 100 columns; run memory, log size and timeout limits apply.
- —Access control is team-wide today; per-client workspaces are in development (Partner Console).
- —No public management API or Git integration for workflows yet — import/export is JSON; versions are built in.
- —No native PSA/RMM pieces (HaloPSA, ConnectWise, Autotask, NinjaOne) yet; use the HTTP step.
Common questions
Is the Workflow Builder no-code?
Yes for the whole path of triggers, steps, branches, loops, agents, documents, approvals and bots. A code step (JavaScript/TypeScript with npm) exists for the cases a formula can't express.
How is it different from Zapier, Make or n8n?
The canvas and connectors are familiar. The difference is the step types: an agent with tools, a document step that returns fields, approvals as steps, and a bot that works on screens without an API — on the same canvas, in the same run log.
Can I test a single step without running the whole flow?
Yes — select the step, click Test, see the output in the test panel; the data selector shows what later steps can use.
How do versions work?
Every publish creates a version; drafts don't affect production; roll back in one click. Step versions don't upgrade on their own.
What does a run cost?
Plans are priced by workflow runs (plus bot runs, pages and conversations for those steps). Steps inside a run, users and integrations are not metered; you bring your own LLM key for agents.
Can another agent trigger my workflow?
Yes, through the MCP tool trigger: the flow becomes a tool that Claude, ChatGPT, Copilot Studio or another ElectroNeek agent can call with structured inputs.